Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
  • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
  • G06F21/31—User authentication
  • G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
  • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
  • G06F21/31—User authentication
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
  • G06F21/60—Protecting data
  • G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
  • G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
  • G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
• • G—PHYSICS
  • G07—CHECKING-DEVICES
  • G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
  • G07C9/00—Individual registration on entry or exit
  • G07C9/30—Individual registration on entry or exit not involving the use of a pass
  • G07C9/32—Individual registration on entry or exit not involving the use of a pass in combination with an identity check
  • G07C9/37—Individual registration on entry or exit not involving the use of a pass in combination with an identity check using biometric data, e.g. fingerprints, iris scans or voice recognition

Definitions

• the invention relates to a method of authentication of a physical object using both a helper data and a control value associated with a reference object; the method including: generating a first property set using information comprising the helper data and a metric associated with the physical object, generating a second property set using a noise compensating mapping on information comprising the first property set, establishing a sufficient match between the physical object and the reference object using the second property set and the first control value.
• the invention further relates to an apparatus for authentication of a physical object using both a helper data and a control value associated with a reference object further comprising: a first generation means arranged to generate a first property set using information comprising the helper data and a metric associated with the physical object, a second generation means arranged to generate a second property set using a noise compensating mapping on information comprising the first property set, a comparing means arranged to generate a sufficient match between the physical object and the reference object using the second property set and the first control value.
• Identification and authentication are commonly used techniques for establishing identity. Identity could be the identity of a person or an object. Prime examples of application areas for identification and authentication are access control for buildings,
• an object with an alleged identity is offered for authentication. Subsequently characteristics of the object offered for authentication are matched with those of the enrolled object with the alleged identity. If a sufficient match is found the identity of the object being authenticated is said to be the alleged identity.
• Authentication thus deals with matching one object being authenticated to one enrolled object based on the alleged identity.
• the authentication process is generally preceded by an enrolment process. During this enrolment characteristics of the object at hand are measured and stored. Based on the measured data so-called template data is generated for the object. This template data is used during the authentication process for matching enrolled objects with the measured characteristics.
• Template data may at first glance present little value. However when this data is used on a regular basis to perform financial transactions its value becomes obvious. Furthermore in case of biometric authentication systems template data may also comprise privacy sensitive biometric data, and therefore have an even greater value. In classical system template data is generally unprotected, and thereby susceptible to malicious attacks.
• helper data and a control value are used for authenticating a physical object. Both are generated during enrolment and are used instead of the actual template data.
• the helper data is generated using the template data, but characteristics of the template data are obfuscated in such a way that there is hardly any correlation between the template data and the helper data.
• the control value is generated in parallel with the helper data and serves as a control value for the authentication process.
• the helper data and control value are used during authentication.
• First the helper data is combined with data acquired from the physical object (e.g. facial feature data).
• the combined data is subsequently "condensed” into a second control value.
• This second control value is matched with the control value generated during enrolment. When these control values match the authentication is said to be successful.
• the authentication process verifies whether the metric obtained from the physical object during authentication sufficiently matches the template data. Assume the physical object is the same as the reference object, the combined data (helper data and metric data) are passed to a noise compensating mapping to compensate for measurement noise in the metric data.
• the noise compensating mapping determines to a large extent whether or not a sufficient match is found between the physical object and the reference object. Consequently the classification boundaries of a helper data system, that determine whether an object matches or not, are primarily determined by characteristics of the noise compensating mapping used. In a conventional helper data system where two objects substantively resemble one another, such that the classification boundaries for the individual objects overlap, it is not possible to differentiate between noise and differentiating features for these objects.
• the method as set forth in the introductory paragraph is characterized in that it further comprises the following steps: generating an error measure by quantifying the noise removed by the noise compensating mapping using the first property set and information derived from the noise compensating mapping, and using said error measure for generating an authentication decision.
• the noise robust mapping is used to provide resilience to measurement errors in the (bio)metric data acquired from the physical object.
• the noise compensating mapping can be interpreted as the inverse of the noise robust mapping, where the noise robust mapping adds noise resilience, the noise compensating mapping uses this to reconstruct the original message in the presence of noise. Provided the noise robust mapping is sufficiently robust, or the measurement noise is sufficiently small, successful authentication is possible. The noise robust mapping effectively determines the classification boundaries of such authentication methods.
• the present invention uses the input and output of the noise compensating mapping to quantify an error measure that can be used to further contract the classification boundaries of the authentication method by specifying additional constraints based on said error measure. By applying such constrains it is possible to further differentiate between objects that otherwise could result in a false positive authentication.
• the same error measure that can be used for shaping the classification boundaries can be used to establish a probability measure indicative of the probability of a false positive.
• a straightforward way to establish such a probability measure is to use the number of symbol errors corrected by the noise robust mapping and divide that by the maximum number of symbol errors that the noise robust mapping can correct. The resulting ratio is indicative of the chance of a false positive.
• This probability measure can also be used for making soft decisions.
• Soft decisions are particularly advantageous in multi-modal authentication systems, where the result from the template protected authentication system is combined with the results of other authentication methods. Combining probabilities instead of bi-exact decisions can contribute significantly to the overall quality of the decision.
• a further beneficial application of the present invention is application in an authentication method in which multiple candidate objects are compared to the metric data, thereby establishing whether there are other objects that present a better match than the alleged identity.
• An alternate advantageous application is to apply the steps as presented for the method of authentication in a method of identification of a physical object.
• Identification can be interpreted as a repeated authentication process, where multiple objects from the set of enrolled objects are matched with the physical object.
• An identification method further requires an additional step to select the best match, or most likely reference object/identity out of all matching reference objects/identities.
• an identification method employing helper data multiple reference objects could be found in the enrolment database whose helper data in conjunction with the (bio)metric Y result in the same second property set. If this is the case a further selection step is needed.
• a noise measure can be established, or a probability measure. These can be used in the final selection step, by selecting the reference object/identity that resulted in the smallest noise measure, or selecting the most likely reference object/identity.
• the present method can also be applied in an advantageous fashion in multimodal identification methods.
• an identification method applying helper data can create a set of matching objects and accompanying probability measures. These probability measures provide additional differentiating information over the actual identification decisions. By combining candidate sets and probability measures from the individual methods that make up the multi-modal identification method it is possible to make a more reliable identification decision.
• the apparatus as set forth in the second paragraph further comprises a third generation means arranged to generate an error measure by quantifying the noise removed by the noise compensating mapping using the first property set and information derived from the noise compensating mapping, and an authentication decision means arranged to use said error measure to generate an authentication decision.
• Fig. 1 is a block diagram of the enrolment and authentication process in an authentication system for a physical object employing template protection according to the prior art.
• Fig. 2 is a graphical representation of soft matching and classification boundary shaping according to the present invention.
• Fig. 3 is a block diagram of an apparatus for authentication of a physical object employing template protection according to the present invention.
• Fig. 4 is a block diagram of an implementation of a subsection of an apparatus for authentication of a physical object employing template protection according to the present invention.
• Fig. 5 is a block diagram of an alternate implementation of a subsection of an apparatus for authentication of a physical object employing template protection according to the present invention.
• Fig. 6 is a block diagram of an alternate implementation of an authentication decision means as used by an apparatus for authentication of a physical object employing template protection according to the present invention .
• the present invention is described primarily for use in authentication systems, the present method can be applied to identification systems in an equally advantageous way.
• a metric obtained from a physical object with an alleged identity is matched with enrolment data associated with a reference object with the alleged identity.
• a metric obtained from a physical object without an alleged identity is matched with enrolment data associated with a series of reference objects to establish an identity. Both processes effectively perform a comparison of a metric obtained during authentication/identification, and compare this with enrolment data of at least one reference object.
• Fig. 1 depicts an enrolment process ENRL on the left hand side, during the enrolment process ENRL a helper data W and a control value V are generated for the object being enrolled. This data is subsequently stored in the authentication data set ADS, located in the middle.
• AUTH depicted on the right hand side, a physical object (not shown in Fig. 1) with an alleged identity is authenticated.
• the authentication data set ADS is searched for a reference object with the alleged identity. If there is no such reference object the authentication will fail. Provided the reference object is found, a first helper data Wl and an accompanying first control value Vl associated with the alleged identity are retrieved from the authentication data set ADS. This data is used to decide whether or not the physical object being authenticated sufficiently matches the reference object, resulting in a positive authentication.
• the helper data system is used to authenticate persons using biometric data in the form of fingerprint data.
• the biometric template data comprises a graphical representation of the lines and ridges of the core area of the fingerprint. Issues such as the orientation and localization of the core area during acquisition are beyond the scope of the present description.
• a person presents his or her finger to a fingerprint scanner.
• the result from one or more fingerprint scans is used to construct a biometric template X.
• a possibly secret property set S is chosen.
• the property set S is mapped onto a property set C by means of a noise robust mapping NRM.
• helper data W is combined with biometric template X to produce a helper data W.
• the property set S and the noise robust mapping NRM are chosen such that the resulting helper data W does exhibit little or no correlation with the biometric template data X.
• the use of helper data W does not expose the biometric template data X to malicious users.
• control value V is generated using the property set S.
• the control value V can be identical to the property set S this is not advisable in systems where security is an issue.
• a cryptographic hash function is a good example of such a one-way mapping. If security is not critical a non oneway mapping could be used.
• the pair of helper data W and control value V are stored in the authentication data set ADS.
• helper data and control value pairs can be generated easily by selecting different property sets S. Multiple helper data and control value pairs can be particularly useful for managing access levels or for system renewal.
• the authentication data set comprises only a single helper data and control value per enrolled object.
• a biometric data Y fingerprint
• a physical object not shown in Fig. 1
• an alleged identity is provided.
• the next step is to check whether the authentication data set ADS contains a first helper data Wl and a first control value Vl for a reference object with said alleged identity. If this is the case the first helper data Wl and the first control value Vl associated with the reference object are retrieved.
• biometric data Y from the physical object is combined with the first helper data Wl resulting in a first property set Cl.
• the biometric data Y can be interpreted as a noisy version of the biometric template X:
• the first helper data Wl can be represented by template data X and property set C:
• the first property set Cl is passed to the noise compensating mapping NCM to produce a second property set Sl .
• the reference object corresponds with the physical object.
• the noise compensating mapping NCM will reconstruct a second property set Sl that is identical to the original property set S as used during enrolment for generating the first helper data Wl.
• the first property set Sl is subsequently used to compute a second control value V2 in a similar fashion as the first control value Vl.
• Next second control value V2 is compared with the first control value Vl generated during enrolment. Provided the noise robust mapping NRM provides sufficient resilience to noise the second control value V2 will be identical to the first control value Vl. If these values are identical, the authentication is successful, and the identity of the physical object is established as being the alleged identity.
• the noise robust mapping NRM can be selected from a wide variety of mappings.
• a simple noise robust mapping NRM could involve the duplication of input symbols.
• the noise compensating mapping NCM would require a majority vote using the received symbols.
• a more elaborate noise robust mapping NRM can be selected such as a Reed Solomon Error Correcting Code encoder.
• Fig. 2 illustrates how the present invention can be used to shape the classification boundaries of an authentication system employing template protection.
• Fig. 2 shows two different domains the S-domain SDOM, and the C- domain CDOM.
• the S-domain SDOM is a two-dimensional projection of a potentially N- dimensional space that holds potential elements of the property set S chosen during the enrolment phase.
• the C-domain CDOM is a two-dimensional projection of a potentially M- dimensional space that holds potential elements corresponding with property set C.
• a property set SVALl is chosen.
• the property set SVALl is used to generate a property set CVALl that in turn is used to generate helper data.
• this helper data will be retrieved and combined with a metric obtained from the object into the first property set Cl .
• the noise compensating mapping will attempt to compensate for that noise using the noise resilience added by the noise robust mapping.
• the region REGl corresponds to all property sets that will be mapped onto SVALl by the noise compensating mapping. Thereby the circle REGl corresponds to the classification boundary for this particular object.
• the regions REG2 and REG3 depicted in Fig. 2 are associated with two other objects OB J2 and OBJ3 respectively (not shown). Again the circles correspond to the classification boundaries used by the system for the respective objects.
• the present method provides a solution to this problem by quantifying an error measure ERR between the first property set Cl generated during authentication and CVAL2 or CVAL3 respectively. Subsequently additional constraints can be added to this error measure ERR for each particular object, thereby contracting the classification boundaries for that particular object as indicated by REG2' and REG3' respectively. By careful selection the classification boundaries can be contracted such that there is no more overlap between the regions.
• classification boundaries can be shaped for each individual object using constraints. These constraints can be unique for each individual object and can be stored in combination with their respective helper data and control value in the authentication data set ADS.
• the error measure ERR and associated constraints can be either scalar or vector based.
• the obvious advantage of using a vector is that multiple individual constraints can be set to vector elements, thereby facilitating more detailed classification boundary shaping.
• the constraints could also comprise thresholds to combinations of vector coefficients thereby modelling relations between vector coefficients.
• the classification boundaries in Fig. 2 are of similar size this is not a prerequisite of the present method.
• the shape and size of the regions are determined primarily by the noise robust mapping NRM and need not be identical for all values in the C- domain CDOM.
• Fig. 3 presents a block diagram of an apparatus for authentication of a physical object employing template protection according to the present invention.
• a first helper data Wl and a control value Vl are obtained from the authentication data set ADS.
• the first generation means GMl generates the first property set Cl using the first helper data W land a metric Y obtained from the physical object (not shown).
• the second generation means GM2 subsequently generates the second property set Sl by applying a noise compensating mapping NCM on the first property set Cl.
• the third generation means GM3 subsequently combines the first property set Cl and the second property set Sl into an error measure ERR.
• the error measure ERR together with the second property set Sl and the control value Vl are inputs to the authentication decision means ADM that generates a decision D and a probability measure P related to that decision.
• Fig. 4, 5, and 6 provide further details with respect to the third generation means GM3 and the authentication decision means ADM.
• the noise robust mapping NRM is at the heart of the present invention. During authentication characteristics of the noise robust mapping NRM are used to help shape the classification boundaries used in the authentication decision means ADM. Error Correcting Code encoders, or ECC encoders are good examples of noise robust mappings.
• ECC encoders are used to add redundancy to messages to facilitate message extraction after operations that inject noise into the message.
• ECC encoders can be classified in a variety of ways the classification used here distinguishes two classes:
• ECC codes where input and output code-words consist of symbols from the same alphabet, input and output have a similar format.
• Systematic ECC codes are members of the first class. When a message is encoded by means of a systematic ECC encoder the message is copied unaltered into the codeword, and the parity bits are effectively appended to the message.
• a systematic ECC decoder in turn maps the input codeword consisting of the concatenated message and parity bits onto a new corrected codeword of the same format.
• a first property set Cl is generated using the first helper data.
• the property set Cl can be seen as the property set C with a superimposed noise component. Provided that the noise component is small enough, or the ECC used robust enough the first property set Cl can be successfully corrected by means of the ECC decoder.
• Establishing an error measure between the input and output code words for members of the first class of ECC codes can be as simple as subtracting the ECC decoder input from the ECC decoder output.
• Fig. 4 shows a block diagram of a subsection of an apparatus according to the present invention that exploits these characteristics. The top part of Fig. 4 depicts a simple third generation means GM3.
• the third generation means GM 3 generates an error measure ERR by subtracting the first property set Cl from the second property set Sl.
• the error measure in turn is sent to a comparator CMP in the authentication decision means ADM.
• This particular authorization decision means ADM performs three steps: 1. It compares the first control value Vl with information derived from the first property set Sl to establish whether the noise compensating mapping NCM was able to reconstruct the property set S used during enrolment.
• the first step corresponds to the step in a conventional helper data system where a sufficient match is established between control value Vl and information derived from second property set Sl .
• control value Vl could be identical to the second property set Sl.
• control value Vl could be generated by means of a one way function, such as a cryptographic hash, in doing so authentication remains possible without exposing the second property set Sl.
• the error measure ERR can be used to further contract the classification boundaries; the second step.
• Error measure ERR constraints can vary with the error measure ERR.
• the error measure ERR is of a scalar nature, such as the number of symbol errors in the first property set Cl, a further restriction could be imposed that requires the number of bit errors to be smaller than a predetermined scalar threshold t.
• a further restriction could be imposed that requires the individual coefficients have to be smaller than accompanying predetermined values, which in turn could be represented using a predetermined threshold vector t.
• the pre-determined threshold t may consist of both threshold values for coefficients, as well as threshold for combinations of vector coefficients, thereby accounting for relations between respective vector coefficients.
• the probability measure P effectively provides additional information that can be used for further processing together with or even instead of the decision D.
• the apparatus in Fig. 3 addresses authentication, but with minor enhancements could be used for identification. In case of identification multiple objects from the authentication data set ADS, are compared with the metric data Y acquired from the physical object . In case of identification the physical object being identified does not provide an alleged identity. Instead the identity of the physical object can be derived from the identity of the reference object that provides a sufficient or best match. To this end the apparatus can be extended with an identity establishing means, that can retrieve the identity of the reference object from the authentication data set ADS, and can, based on the decision D and or the probability P, establish the identity of the physical object to be identical to that of the reference object. Fig.
• the number of symbol errors corrected by the noise compensating mapping NCM could be used as an indication of a probability measure. If needed this particular measure can be normalized by dividing through the maximum number of errors that the noise robust mapping can compensate.
• the probability measure P that is based on the error measure ERR is only valid when the error measure ERR is valid, that is when the first step of the authentication decision means ADM is successful.
• the three steps described here are representative of an embodiment according to the invention, but not limiting.
• the individual steps for generating a decision D and the accompanying probability measure P can be combined.
• a simple way to combine these steps would be to establish a probability measure, e.g. through careful weighing of errors of individual vector coefficients, and subsequently using the probability in conjunction with a pre-determined threshold to determine whether the authentication decision should be positive or negative.
• the apparatus depicted in Fig. 5 resembles that in Fig. 4 but includes an alternate implementation of a third generation means GM3 and an authentication decision means ADM.
• the key difference with respect to the third generation means GM3 being that the implementation of the third generation means GM3 in Fig. 5 can be used for error correcting codes of both classes of ECC as mentioned earlier.
• the third generation means GM3 comprises a fourth generation means GM4 that generates a third property set C2 by applying the noise robust mapping NRM on the second property set Sl. Regardless of the requirements of the input and output of the noise robust mapping NRM, the third property set C2 has the same format as that of the first property set Cl . Subsequently a fifth generation means GM5 can subtract the first property set Cl from the third property set C2 to obtain an error measure ERR for both classes of ECC codes.
• Fig. 5 also depicts how a secure apparatus for authentication of a physical object employing template protection can generate a second control value V2 from the first property set Sl using a one-way function h.
• the present invention can also be used for generating soft-decisions.
• a system applying soft- decisions capitalizes on the fact that the probability measure P comprises more information than the decision D.
• This additional information can be used in multi-modal identification systems. Instead of determining a single matching object, or a list of matching objects it is possible to provide a list of matching objects with probabilities. These probabilities can be used as differentiating factors when combining the results with the other parts of the multi-modal system.Due to the additional information/resolution provided by the probability measures the identification process can be made more reliable.
• Fig. 6 depicts an implementation of an authentication decision means ADM that can be used in an embodiment of the present invention.
• the authentication decision means ADM uses a two-stage approach, in the first stage an interim decision ID is established based on the a control value Vl and a second control value V2 derived from the second property set Sl .
• the interim decision ID is subsequently passed to an evaluation means EVM, where subsequently the error measure ERR is compared with a predetermined threshold t. Based on this comparison the evaluation means EVM can subsequently reject or accept a positive interim decision ID to obtain a decision D.
• the present invention was illustrated using examples that apply a single first helper data Wl and a first control value Vl, the present invention can be used advantageously in systems that apply multiple pairs of helper data and control values to authenticate physical objects.
• additional constraints based on the error measure ERR only allow contraction of classification boundaries.
• the use of multiple pairs of helper data and control values allow the extension of classification boundaries, and thereby facilitates true shaping, contraction and extension, of the classification boundaries for objects.
• any reference signs placed between parentheses shall not be construed as limiting the claim.
• the word “comprising” does not exclude the presence of elements or steps other than those listed in a claim.
• the word “a” or “an” preceding an element does not exclude the presence of a plurality of such elements.
• the invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer.
• the device claim enumerating several means several of these means can be embodied by one and the same item of hardware.
• the mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Computer Security & Cryptography (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Software Systems (AREA)
• Computer Hardware Design (AREA)
• General Engineering & Computer Science (AREA)
• Health & Medical Sciences (AREA)
• Bioethics (AREA)
• General Health & Medical Sciences (AREA)
• Human Computer Interaction (AREA)
• Medical Informatics (AREA)
• Databases & Information Systems (AREA)
• Collating Specific Patterns (AREA)

Priority Applications (2)

Applications Claiming Priority (3)

Related Child Applications (1)

Publications (1)

Family

ID=37482040

Family Applications (2)

Family Applications Before (1)

Country Status (5)

Families Citing this family (5)

Family Cites Families (15)

• 2006
  • 2006-05-23 CN CNA200680019152XA patent/CN101185104A/zh active Pending
  • 2006-05-23 WO PCT/IB2006/051646 patent/WO2006129241A2/en not_active Application Discontinuation
  • 2006-05-23 EP EP09008662.0A patent/EP2159759B1/en active Active
  • 2006-05-23 US US11/916,094 patent/US8122260B2/en active Active
  • 2006-05-23 EP EP06765706A patent/EP1891608A2/en not_active Withdrawn
  • 2006-05-23 JP JP2008514259A patent/JP2008542898A/ja active Pending

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events